NeuralStackAI Law Radar’s newest update paints a picture of AI regulation that is becoming more concrete, more sector-specific and more focused on particular risky uses of AI rather than broad, omnibus statutes. In the United Kingdom, the service now treats the AI (Regulation) Bill [HL] as unlikely to advance in its current form, upgrading its confidence in that assessment after checking against Parliament’s Bills API and recent government policy. At the same time, the register is being expanded with detailed, binding obligations that are already reshaping how automated decision-making and synthetic imagery can be used in practice.
The most substantive additions sit in the UK’s Data (Use and Access) Act 2025, which quietly rewires the country’s approach to automated decision-making under the UK GDPR. Where the old default centered on a broad prohibition in Article 22, the new ss.22A-22D, in force since early February 2026, replace that blunt ban with a more procedural set of duties: notification when automated decisions are used, representation rights, human review and contest mechanisms. Rather than blocking automated decisions outright, the framework pushes companies toward explainability and redress, with AI Law Radar flagging this shift as a newly tracked obligation.
On synthetic imagery, the UK’s DUAA 2025 s.138 marks a sharp turn toward criminalisation of non-consensual intimate deepfakes. In force since February 2026, the provision adds new sections to the Sexual Offences Act 2003 that make it a crime to create or even request a non-consensual intimate deepfake, regardless of whether the content is ever distributed. The tracker originally scoped this offence to AI deployers, but now clarifies that it applies to any person in the UK, aligning the register with the underlying statute and distinguishing these end-user offences from separate tool-supplier offences created later in the Crime and Policing Act 2026.
Algorithmic price-setting is also moving squarely into regulators’ crosshairs. AI Law Radar has added Illinois SB 343 to its register, describing it as an amendment to the Illinois Antitrust Act that would ban algorithmic coordination of rental prices and tracking it as a proposed law awaiting the governor’s signature, with no effective date yet. In a separate correction, the service reclassified SB 343, New Jersey’s FAIR Act and Maryland’s Protection From Predatory Pricing Act under a clearer "Prohibited AI practices" theme, emphasizing that these measures do not merely regulate risk but outright forbid certain algorithmic uses in rent-setting and pricing.
China appears in the changelog via the CAC, NDRC and MIIT’s AI Agents Implementation Opinions, issued in May 2026 and now captured as a distinct obligation in the tracker. While AI Law Radar’s entry is spare on operational detail, the inclusion signals that anthropomorphic or agent-like AI systems are drawing attention from multiple economic and technology regulators in Beijing. By treating these Opinions as a coverage gap worth closing, the tracker underscores how governance of AI "agents" is rapidly becoming a global concern alongside more familiar issues like automated decision-making and synthetic sexual imagery.
Alongside new entries, AI Law Radar continues to refine the accuracy of its existing register, reinforcing that the devil in AI law is often in the citations. The changelog notes a correction to New York’s RAISE Act, clarifying that the enacted frontier AI safety law is S6953-B/A6453-B rather than an earlier cited bill number, while confirming that its future effective date remains unchanged. In Australia, the service fixes an earlier mis-citation of automated decision-making transparency duties, confirming that the Privacy Act amendments actually insert new obligations as APP 1.7-1.9, not APP 1.3, with the commencement date intact.
Other updates point to how governments are pacing enforcement. AI Law Radar reports that South Korea has now confirmed at least a one-year fines grace period for high-impact AI duties under its AI Basic Act, starting from the law’s January 2026 effective date. In the United States, the tracker now treats Georgia’s SB 540 as having a firm effective date in mid-2027 based on an official Senate press release, and resolves uncertainty around Rhode Island’s Healthcare AI Documentation Act, marking it effective upon passage when the governor signed it in June 2026. These entries focus less on new obligations and more on calibrating when penalties and compliance clocks actually start ticking.
Tennessee’s experience shows how fast-moving political negotiations can transform AI bills between introduction and enactment. AI Law Radar describes SB 1700 as originally framed around companion-chatbot safety, but notes that Senate amendments adopted in April 2026 stripped those operative restrictions before passage. The law that ultimately emerged, Public Chapter 1082, now functions only as a study mandate directing a state commission to examine potential AI and chatbot regulation, with no report deadline or compliance requirements. For companies tracking regulatory risk, the distinction between an enforceable safety duty and an open-ended study can be the difference between urgent remediation work and longer-term policy watchfulness.
Why this matters
Viewed together, the changes in AI Law Radar’s register show that AI governance is moving away from single flagship acts and toward a dense patchwork of targeted obligations, prohibitions and grace periods. The UK’s pivot from an unlikely AI (Regulation) Bill to an AI Growth Lab and sector-sandbox strategy, coupled with granular DUAA and GDPR reforms, illustrates a preference for modifying existing data and criminal frameworks rather than building a standalone AI statute. In the US, state-level bills about rental algorithms, healthcare AI documentation and conversational AI safety, plus study mandates like Tennessee’s, highlight a fragmented landscape where legal risk depends heavily on which state a product touches.
For AI builders and deployers, the practical takeaway is that compliance can no longer be addressed as a monolithic "AI law" problem. Developers of decision-support systems must absorb detailed notification and human-review duties in the UK, while anyone experimenting with synthetic intimate imagery faces criminal exposure even at the request stage. Rental platforms leaning on algorithmic coordination need to scrutinize antitrust amendments in Illinois and other states, and designers of anthropomorphic AI agents have to be aware that Chinese authorities are now issuing implementation guidance specific to their products. Across jurisdictions, the fine print—bill numbers, clause locations, scope corrections—will increasingly determine which AI features are legally viable and which cross into prohibited practice.
The forward trajectory suggested by AI Law Radar’s changelog is one of more refinement, more cross-references to existing legal systems, and more attention to enforcement timing. As grace periods in places like South Korea roll forward and future effective dates in Georgia, New York and Australia approach, companies will need to shift from tracking to implementation, turning regulatory intelligence into concrete changes in design and deployment. With sector sandboxes, study mandates and implementation opinions proliferating alongside criminal offences and antitrust bans, the next phase of AI regulation is likely to be defined less by new acronyms and more by how fast organisations can parse, align with and, where necessary, challenge an increasingly granular rulebook.